昨日阅读1小时，累计阅读45小时。

昨天阅读1小时，累计阅读233小时。

昨天阅读1小时，累计阅读193小时。

昨日阅读1小时，累计阅读234小时。

昨天阅读1个小时，累积阅读52个小时。

And when a blowup happens, they invoke uncertainty, something called a Black Swan (a high-impact unexpected event), after a book by a (very) stubborn fellow, not realizing that one should not mess with a system if the results are fraught with uncertainty, or, more generally, should avoid engaging in an action with a big downside if one has no idea of the outcomes. What is crucial here is that the downside doesn’t affect the interventionist. He continues his practice from the comfort of his thermally regulated suburban house with a two-car garage, a dog, and a small play area with pesticide-free grass for his overprotected 2.2 children.
Imagine people with similar mental handicaps, people who don’t understand asymmetry, piloting planes. Incompetent pilots, those who cannot learn from experience, or don’t mind taking risks they don’t understand, may kill many. But they will themselves end up at the bottom of, say, the Bermuda Triangle, and cease to represent a threat to others and mankind. Not here.
So we end up populating what we call the intelligentsia with people who are delusional, literally mentally deranged, simply because they never have to pay for the consequences of their actions, repeating modernist slogans stripped of all depth (for instance, they keep using the term “democracy” while encouraging headcutters; democracy is something they read about in graduate studies). In general, when you hear someone invoking abstract modernistic notions, you can assume that they got some education (but not enough, or in the wrong discipline) and have too little accountability.
Now some innocent people—Ezidis, Christian minorities in the Near (and Middle) East, Mandeans, Syrians, Iraqis, and Libyans—had to pay a price for the mistakes of these interventionist as currently sitting in comfortable air-conditioned offices. This, we will see, violates the very notion of justice from its prebiblical, Babylonian inception—as well as the ethical structure, that underlying matrix thanks to which humanity has survived.
The principle of intervention, like that of healers, is first do no harm (primum non nocere); even more, we will argue, those who don’t take risks should never be involved in making decisions.

昨天阅读5小时，累计阅读70小时。

昨日阅读2小时，累计阅读886小时      
挑战第三百六十六天   读25页书，完成当日任务

昨日阅读1小时，累积阅读205小时

昨日阅读2小时，累计阅读758小时

看看看

今天阅读1小时，持续每天阅读累计149小时。

昨日阅读2小时，累计阅读1139小时。

昨日阅读3小时，总计116.5 小时

机器学习导论-2nd Edition

1.2.2 分类
在信用评分（credit scoring）（Hand 1998）中，银行要计算在给定信贷额度和客户信息情况下的风险。
客户信息包括我们已经获取的数据以及与计算客户财力相关的数据，即收入、存款、担保、职业、年龄、以往经济记录等。
通过这些特定的数据，我们可以推断出一般规则，表示客户属性及其风险性的关联性。
即
机器学习系统用一个模型来拟合过去的数据，以便能够对新的申请计算风险，从而决定接受或拒绝该项申请。

这是一个分类（classification）问题的例子，有两个类：低风险客户和高风险客户。
客户信息作为分类器的输入（input），分类器的任务是将输入指派到其中的一个类

利用以往数据进行训练后，学习得到的规则可能具有如下形式：
        IF  income> θ1 AND savings > θ2 THEN low-risk   ELSE high-risk
其中θ1和θ2是合适的值。这是判别式（discriminant）的一个例子，它是将不同类的样本分开的函数。

这样的规则主要用途是预测（prediction）：一旦拥有了拟合以往数据的规则，如果未来与过去类似，就能够对新的实例做出正确的预测。

某些情况下，可能不希望做0、1类型的判断，而是希望计算一个概率值P（Y|X），其中X是顾客属性，Y是0或1，分别代表低风险和高风险。
从这个角度看，可以将分类看作学习从X到Y的关联性。于是，给定X=x，如果有P（Y=1|X=x）=0.8，则我们就说该客户为高风险得可能性为80%，或说低风险的可能性有20%
我们可以根据可能的收益和损失来决定接受或拒绝这笔贷款业务。

机器学习在模式识别（Pattern Recognition）方面有很多应用。
如：光学字符识别（optical character recognition, OCR）即从字符图像识别字符编码。这是一个多类问题的例子，类与我们想要识别的字符一样多。特别有趣的是手写体字符的识别问题。
我们没有字符“A"的形式化描述，涵盖所有“A”而不涵盖任何非“A"，没有这种形式化描述，就要从书写者那里取样，从实例中学习关于“A”的定义。尽管我们不知道是什么因素使得一个图像被识别为“A”，但是我们确信所有这些不同的“A”的图像都具有某些共同的特征，这正是我们希望从实例中提取的。

昨日阅读1小时，累计阅读150小时

昨日阅读0.5小时，累计阅读374.5小时

昨天阅读2小时, 累计阅读430小时
阅读2小时-O Done.  完成

Material: Michael E.Porter ‘s paper named a Strategy for Health Care Reform- toward a value based system issued in Jul.09,2009 the New England Journal of Medicine.
Reflection: what we need now is a clear national comprehensive vision for the kind of health care system we want to achieve and a path for getting there. The central focus shall be increasing value for patients –the health outcomes achieved per dollar spent. Good outcomes that are achieved efficiently are the goal, not the false savings from cost shifting and restricted service. The only way to truly contain costs in health care is to improve outcomes: in a value-based system, achieving and maintaining good health is inherently less costly than dealing with poor health. True reform will require both moving toward universal insurance coverage and restructuring the care delivery system. How can we achieve universal coverage in a way that will support, rather than impede, a fundamental reorientation of the delivery system around value for patients? There are several critical steps. First, we must change the nature of health insurance competition. We must introduce regulation to end coverage and price discrimination based on health risks or existing health problems. In addition, health plan shall be required to measure and report their subscriber’s health outcomes, starting with a group of important medical conditions. Second, we must keep employers in the insurance system; third, we need to address the unfair burden on people who has no access to employers-based coverage who therefore face higher premiums and greater difficulty on securing coverage. Fourth, to make individual insurance affordable, we need large statewide or multistate insurance pools, fifth, income-based subsidies will be needed to help lower income people to buy insurance, finally once a value-based insurance market had been established, everyone must be required to purchase the insurance so that young and healthier people could not opt out. For restructuring the delivery system where most of the value is created and most of the costs are incurred. First, measurement and dissemination of health outcomes shall become mandatory for every provider and every medical conditions. Second, we need to radically reexamine how to organize the delivery of prevention, wellness, screening and routine health maintenance services. Third, we need to reorganize care delivery around medical condition. Fourth, we need a reimbursement system that aligns everyone’s interests around improving value for patients. Fifth, we must expect and require providers to compete for patients, based on value at the medical-condition level, both within and across state borders. Sixth, electronic medical records will enable value improvement, but only if they support integrated care and outcome measurement. Finally consumers must become much more involved in their health and health care. Michael port did not mention why this is the right strategy.

昨天阅读1小时，累计阅读1215小时。

昨日阅读1小时，累计阅读8小时

https://bbs.pinggu.org/forum.php? ... =view&ctid=2638
请每日在最新充实计划主题帖回复参与活动,方可领取奖励

昨天阅读1小时，累计阅读41小时。

昨日阅读4小时，累计阅读44小时。

https://bbs.pinggu.org/forum.php? ... =view&ctid=2638
请每日在最新充实计划主题帖回复参与活动,方可领取奖励

谢谢楼主分享！！！